Flow-Aware Ellipsoidal Filtration for Persistent Homology of Recurrent Signals

Description

This invited talk presents recent work on flow-aware ellipsoidal filtrations for persistent homology of recurrent signals, based on our publication in Chaos (DOI: 10.1063/5.0317749). The method introduces a spatio-temporal covariance framework to estimate local flow geometry and construct anisotropic neighbourhoods that capture directional and temporal structure in the data. Compared to isotropic approaches such as Vietoris--Rips filtrations, the proposed method improves topology-preserving denoising and first-return-time estimation. The talk will be delivered at the 15th AIMS Conference, held in Athens, Greece, 6--10 July 2026.

Period	6 Jul 2026 → 10 Jul 2026
Held at	National and Kapodistrian University of Athens, Greece
Degree of Recognition	International